Three Things Broke Today That The Dow Didn't Notice

OpenAI just told its own people it can't pay for the data centers it ordered. The U.A.E. walked out on OPEC. A small biotech in California nearly doubled survival on the worst cancer there is. The Dow shed 25 points and the tape thinks nothing happened today. The tape is wrong.

First, the closes — because we have to start there. S&P 500 7,138.80, down half a percent. Nasdaq 24,663.80, down nine-tenths. Dow 49,141.93, basically flat. WTI ripped to $99.93 — up $3.56 in a session — and Brent touched $111.26, the highest close since the cease-fire. Gold finally took a breath, down $83.90 to $4,591.50. 10-year at 4.352%, 2-year 3.843%. The Fed announces tomorrow at 2pm ET. Mag-7 reports tonight (Alphabet, Microsoft, Amazon, Meta) and Apple after the close Thursday. Now — the three things.

1. OpenAI just told its own people it can't pay the bills

This is the lede on the front page (Berber Jin reporting). OpenAI is missing its own user and revenue targets ahead of an IPO they want to do this year. The CFO, Sarah Friar, has been telling other leaders inside the company that she's worried they won't be able to pay for future computing contracts if revenue doesn't accelerate. The board is now scrutinizing every new data-center deal Sam Altman wants to sign. Read that twice: the directors are slowing him down.

The market figured this out fast. Oracle −4% — they have a $300 billion compute deal with OpenAI. CoreWeave −5.8% — that's their biggest cloud partner. SoftBank −9% in Tokyo, the worst day since November. AMD −3.4%. The whole "AI infrastructure" trade got hit, but the AI builders — Microsoft, Google, Nvidia, Broadcom — mostly held up. That tells you something. The market is finally separating the people who own the AI revenue from the people who are financing the AI buildout. Those are not the same trade anymore.

And read this with it — Asa Fitch's Heard on the Street column on B14. Microsoft and OpenAI quietly redrew their partnership Monday. Second amendment in six months. OpenAI is now free to sell tools outside Azure; the revenue-share through 2030 is capped; Microsoft has started integrating Anthropic models alongside OpenAI's. Translation: this is a slow split. The "couple" everyone thought was inseparable is moving toward separate apartments.

What I'm doing in the book — I'm trimming ORCL, CRWV, AMD inside the AI sleeve (which sits at 18% of the aggressive book after last weekend's reset). I'm not touching NVDA, GOOGL, MSFT, AVGO. Those are owners. The trim names are counterparties. Microsoft actually looks better after this news, not worse, because their diversification away from OpenAI exclusivity is a feature now. We'll see what the prints do tonight.

2. The U.A.E. walked out on OPEC

Front-page lede on the right side (Said, Kantchev, Feng, Faucon). U.A.E. announced it's leaving OPEC. This is the most consequential oil-cartel news in thirty years and the editorial board (A14) frames it cleanly: U.S. fracking plus the Iran war just fractured the cartel that controlled global oil prices for sixty years.

The numbers tell you why. U.S. break-even on a fracked barrel is around $50. OPEC's budgets balance somewhere between $80 and $120. Abu Dhabi has 1.4 million barrels a day of capacity sitting unused above its OPEC quota. More than half its exports route outside the Strait of Hormuz, so it doesn't even need OPEC discipline to stay liquid through the war. Iran got an exemption. Iraq, Venezuela, and Libya are essentially exempted by sanctions or strife. Saudi Arabia and the U.A.E. were the only two producers actually carrying the cuts — and the U.A.E. just decided it was tired of carrying them.

Oil response was immediate. WTI +$3.56 to $99.93. Brent $111.26. Energy was the only S&P sector up materially today (+1.7%). Even after the Strait reopens — whenever that is — we still need every barrel of UAE/Saudi spare capacity to fill the Iraq and Kuwait holes. That's the editorial board's read and I agree with it.

For the book — I'm holding Energy at 20% (we lifted there on Apr 25). Reinforcing XOM, CVX, COP, HAL, SLB. Adding EOG at 1.5% — pure Permian/Eagle Ford play, cleanest expression of "U.S. fracker won the structural fight." On gold, today's $83 pullback is the first material breath in five weeks. I'm not trimming until I see two consecutive down days; UAE/OPEC and Hormuz both keep the premium intact.

3. Revolution Medicines just delivered the cancer breakthrough of the decade

This one I'm genuinely excited about. David Wainer's column on B14. Revolution Medicines (ticker RVMD) is a small-cap biotech that's been working on RAS — the protein that drives most pancreatic cancers and a big chunk of lung and colorectal too. Scientists have tried to drug RAS for forty years and failed. The protein has no obvious place for a small molecule to grip onto.

RevMed's pill, daraxonrasib, found a way around it. They use a "molecular glue" approach — the drug binds to a separate protein and uses the combined surface to grab RAS. In a late-stage pancreatic-cancer trial, daraxonrasib nearly doubled survival compared to chemotherapy. Pancreatic cancer is one of the most lethal cancers there is. People diagnosed with it usually have months. The FDA is expected to approve later this year.

Now the deal math. RevMed had talks with Merck and AbbVie last year at a $30 billion valuation; no deal closed. The stock has rallied to roughly that market cap, which means an acquirer would now have to pay $40 billion or more. RevMed just raised $2 billion in equity and convertibles, which signals they're funding the standalone path — the Vertex Pharmaceuticals model. Vertex was a perpetual takeover candidate ten years ago; they built a dominant cystic-fibrosis franchise and they're worth $100 billion now.

And here's the kicker. The closest competitor — Erasca — collapsed 48% Tuesday after a patient in their RAS trial died of pneumonia. RevMed is also alleging that Erasca's drug infringes RevMed patents. So overnight, RevMed went from "one of two RAS franchises" to "the only credible RAS franchise."

I'm adding RVMD at 1.5% inside the Healthcare sleeve. The downside protection is real — M&A optionality from Merck (Keytruda goes off patent and they need to fill that hole), Lilly (oncology gap), or J&J. Even on a standalone DCF with peak pancreatic sales of $5-7 billion and modest pipeline credit, I get to about 60% upside. Avoiding ERAS entirely — safety setback plus IP overhang. Watching MRK for the M&A optionality.

4. The upper-middle consumer is starting to crack

This is the quiet story today and it might matter most. Two pieces in today's paper line up on the same point.

The Renaissance Macro chart on A1 shows that durable-goods spending is falling fastest exactly where prices are rising fastest: phones, recreational books, household appliances, new cars, furniture. People aren't paying the higher prices — they're walking away. Healthcare services are still up because that's not optional.

And then Phil Gramm and Michael Solon on the op-ed page run the tariff scoreboard. The One Big Beautiful Bill saved taxpayers $188 billion in 2025. The new tariffs collected $195 billion. We're already net negative. The CBO's projection for 2026 is worse: $331 billion in tariffs vs. $230 billion in tax cuts. Tariff drag exceeds tax-cut benefit by 44%. The NY Fed has measured it — 100% pass-through to import prices. Foreign suppliers aren't eating it. Americans are.

Spencer Jakab's Heard on the Street column (also B14) has the punctuation mark. Travel + Leisure — the timeshare operator — warned about "some move in early-stage delinquencies" on recent loans. Stock down 14%. Marriott Vacations down 8%. Hilton Grand Vacations down 4%. The average timeshare is $23,000, financed at 13%. This is exactly the upper-middle cohort — people with houses and 401(k)s but no private bankers — and they're starting to skip payments. First time the data has moved in five years.

And on top of it: Ares Capital reported this morning that nonaccrual loans in their private-credit fund went from 1.8% to 2.1%. Symplr Software took a $58 million write-down. Cornerstone OnDemand took $63 million. Barclays just took $1.1 billion in credit impairments — biggest quarter since Covid. Credit cycle is turning. Quietly. Before the headline indices catch on.

For the book — reversing my Apr 25 trim of KO. Coke popped 3.9% today on a Q1 beat (biggest one-day since October) and the message is clear: defensive staples are bid as the cyclical narrative cracks. Rebuilding Defensive 7% → 8%. Reinforcing WMT and COST on trade-down. Watching TNL and ARCC; not buying the dip yet, but they're now alarms on the dashboard.

5. King Charles spoke to Congress yesterday and you should pay attention

I usually skip royal news. Not this one. King Charles III — second British monarch in history to address a joint session of Congress — gave a speech that was rubber-stamped by the U.K. government but largely written by him. It was a defense speech. The biggest sustained increase in U.K. defense spending since the Cold War. F-35 joint production with us. AUKUS — "the most ambitious submarine program in history." Article 5. Thousands of U.S. service personnel in the U.K., British troops in thirty American states.

Pair it with John Lehman and Anand Toprani's op-ed on the same page. Lehman was Reagan's Navy Secretary. He and Toprani lay out exactly what made the Reagan defense buildup work — and it wasn't just the dollar figure. It was procurement reform, strategic doctrine, allied force multipliers, fixed-price contracts. The argument: 5% of GDP isn't enough on its own, you need the institutional discipline behind it. We're closer to the beginning of competition with China than the end.

What both pieces tell me: defense isn't a 2026 budget headline. It's a 5-to-7 year backlog story. Holding Defense at 18%. Reinforcing LMT (F-35), GD (AUKUS subs — they report tomorrow), RTX (Patriot/Tomahawk), NOC (B-21/Sentinel), AVAV and KTOS (low-cost-mass / drones). And adding AME — Ametek — at 1.5%. They're acquiring Indicor from CD&R for $5 billion; that broadens our aerospace/defense exposure with industrial instrumentation we'd been wanting in the book.

6. Tonight matters — the Mag-7 prints

Three things I'm watching tonight (Alphabet, Microsoft, Amazon, Meta) and tomorrow (Apple).

Capex guides. If any of them raise their AI infrastructure capex, today's OpenAI shortfall narrative gets blunted and Oracle/CoreWeave reverse some of these losses. If all four hold or trim — the AI infra trade rolls over for real. That's the swing factor.

Margin commentary. AI revenue should start showing through in Microsoft Azure and Google Cloud. If absolute revenue is up but margin compression continues from severance and restructuring (the Apr 25 META keystroke-tracking story), the productivity-via-displacement narrative is right but the payoff is later than 2026.

Apple AI services revenue. Smaller signal but it's the cleanest read on whether consumers will actually pay for AI features.

I'm holding NVDA, AVGO, GOOGL, MSFT through tonight. The bear case — the only thing that would make me trim further — is Microsoft Azure capex slowing and OpenAI dependency reaffirmed. That's the ugly combo. I don't think we get it. We'll see.

The book — where I land

What I think happens next

Three forecasts. I'll show my work.

Energy stays bid into July. The U.A.E. doesn't walk back into OPEC easily — rejoining requires consensus from the same members it just rifted with. Combine that with a 9-month Iraq recovery window and the Iran war drag, and $90+ WTI is the new base case through summer. EOG, XOM, COP win.

The AI infrastructure trade rolls another 5-10% lower into August. If Microsoft and Alphabet hold or trim capex tonight, the Oracle/CoreWeave/SoftBank price action triggers fund-flow rotation away from AI infra and into AI builders (NVDA, GOOGL, MSFT) and verticals (defense, energy, healthcare). The Microsoft/OpenAI partnership amendment is the cleanest signal yet that the partnership is coming apart at the seams.

The upper-middle consumer cracks open in Q3. Tariff drag exceeds tax-cut benefit by 44% per CBO. TNL/Marriott/Hilton timeshare data is the first warning. ARCC nonaccrual ticking. Watching May and June durable-goods data. If confirmed, I'm accelerating WMT/COST reinforce, trimming residual Discretionary, and looking hard at long-duration Treasuries (TLT) as a recession hedge.